Sei sulla pagina 1di 3

DOI: 10.1002/cctc.

201100446

Rational Band Gap Engineering of WO3 Photocatalyst for Visible light Water
Splitting
Fenggong Wang, Cristiana Di Valentin,* and Gianfranco Pacchioni[a]
Owing to its optical, chemical and electronic properties, tungsten oxide (WO3) has direct applications in catalysis, sensing,
electro- or photo-chromic devices, photocatalysis, etc. Photoelectrochemical water splitting is obtained by using WO3 as
a component in Z-Scheme systems devoted to O2 evolution. A
co-catalyst (commonly Pt) improves the efficiency of such systems.[1] The system works under irradiation with l = 450 nm
visible (vis) light thanks to the band gap of WO3, which is
smaller than 3.0 eV. The main drawback is that WO3 is inactive
as regards H2 evolution, since the photoexcited electrons do
not have the right potential required to reduce protons to hydrogen. Indeed, the bottom of the conduction band (CB) is
too low with respect to the H + /H2 redox potential. If this limitation could be removed, WO3 would become an ideal vis-light
photocatalyst for water splitting, eventually in combination
with a co-catalyst. The real challenge is to find a way to upshift both the bottom of the CB and the top of the valence
band (VB), so as to keep the band gap essentially unchanged
for vis-light activity.
Band-gap engineering is possible in oxides via doping or
nanostructuring.[2] Sometimes doping results in localized states
in the gap rather than in shifts of band edges, with negative
consequences on the photocatalytic activity, which can be attributed to the occurrence of recombination processes.[3] In
these cases, the formation of localized impurity states can be
avoided with relatively high dopant concentrations (e.g.
10 %).[4] Electronic structure theory provides a useful complement to experiments for the design of photoactive materials.
Ideally, one would like to identify a single dopant, which, once
incorporated in WO3, can induce the desired upward shifts of
both VB and CB edges. Starting from simple concepts corroborated by DFT calculations based on hybrid functionals, we have
found that Hf substituting W atoms in the g-monoclinic room
temperature WO3 phase at concentrations of  12 % can
induce the desired effects. Despite several attempts to improve
the photoactivity of WO3 by selective doping,[512] we are not
aware of any report on successful doping by Hf. We have identified the main factors controlling the electronic structure of
the material and we give a rational of the doping effect, thus
providing a useful conceptual framework to direct the synthesis of new WO3-based photocatalysts. We will show in particular that the anisotropic structural properties of g-monoclinic

[a] Dr. F. Wang, Dr. C. Di Valentin, Prof. G. Pacchioni


Dipartimento di Scienza dei Materiali
Universit di Milano-Bicocca
Via R. Cozzi 53, 20125 Milano (Italy)
E-mail: cristiana.divalentin@mater.unimib.it
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201100446.

476

WO3 play an important role in determining the response of the


material to structural and electronic modifications.[13]
The g-monoclinic structure of WO3 results from a slight distortion of the cubic structure, with the corner-sharing WO6 octahedra tilted, as shown in Figure 1 a. One can look at it as
a 2  2  2 superstructure based on an idealized cubic unit cell,
or, alternatively as consisting of pseudo-one-dimensional
weakly interacting -W-O-W- chains. The computed direct band
gap is 3.10 eV; experimentally, values in the range 2.63.0 eV
have been reported.[1] The upper part of the VB is mainly derived from the O 2p states. The bottom of the CB is mainly
composed of W 5d states partly mixed in with O 2p orbitals.
Owing to a distorted-octahedral coordination, the W 5d states
split into t2g and eg-like components (see Figure 2). As the W
O bond lengths are asymmetric along both y and z directions
while are nearly symmetric along the x direction,[13] the lower
energy t2g states further split into dyz and dxz/dxy components.
Thus, it can be expected that a distortion of the WO6 octahe-

Figure 1. a) The atomic structures of undoped WO3. b) Schematic representation of the plane effect in HfxW1 xO3. Group A O atoms (xy-plane: red
dashed lines and xz-plane: green dashed lines) are represented by yellow
and blue spheres. The large grey and brown spheres represent W and Hf
atoms, respectively. All small spheres represent O atoms.

 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ChemCatChem 2012, 4, 476 478

Figure 2. The schematic representation of undoped WO3 electronic energy


levels starting from atomic orbitals. The solid and dashed lines represent the
large and small contributions, respectively.

dron and in particular a change in W O bond lengths can


affect both the splitting of t2g states as well as the position of
the CB bottom.
Replacing one lattice W atom in WO3 with Hf has 1) a size
and 2) an electronic effect. The size effect is that Hf4 + , with
a radius of 0.78 , is larger than W6 + (0.62 ), induces some
strain and atomic rearrangement. In particular, four O neighbors to Hf move outwards (0.08, 0.09, 0.35, 0.35 ), while the
other two move inwards (0.12 ), reducing the asymmetric distortion of the HfO6 octahedron. Substitution with Hf also modifies the symmetry and bond lengths of the nearby W ions in
the supercell (see the Supporting Information), therefore
changing the d-states splitting and thus the position of the
dxy/dxz states which make up the bottom of the CB. These distortions result in a moderate up-shift of the CB minimum
(CBM) by 0.23 eV, thus improving the reduction potential of
the material. The shift is not attributed to a direct contribution
of the Hf 5d states which are higher in energy (see Figure 3
and the Supporting Information).
The electronic effect is that Hf has two valence electrons
less than W and induces the presence of two holes in the O 2p
valence band. Two empty O 2p states appear at the center of

Figure 3. The TDOS and PDOS of HfxW1 xO3. The inset is the magnification of
the DOS near CBM.

ChemCatChem 2012, 4, 476 478

the gap (see Figure 3), and the two holes are localized on two
O atoms adjacent to Hf; however, the top of the VB does not
move. The analysis of the total and projected density of states
(TDOS and PDOS) of HfxW1 xO3 (see Figure 3), allows us to rationalize why. The O atoms can be classified into two groups
depending on the atomic plane they belong to: Group A) the
O atoms in the xy- and xz-planes, which contain no Hf; Group
B) the O atoms in planes containing Hf (see Figure 1 b). The
top of the VB is basically made up of the O 2p states of
Group A with the states of Group B O atoms at lower energies
(see the Supporting Information), indicating an electronic
plane effect. As the O atoms of Group A are not directly involved in the bonding with Hf, the position of the top of the
VB is unchanged. Similarly, the bottom of the CB is essentially
made up of 5d states from W atoms lying in the same xyplanes as for Group A O atoms. All W 5d states are found to be
shifted upwards by the presence of Hf, but those in the xyplanes non-containing Hf are the least affected being the farthest apart from the impurity (lattice parameter c > b > a). The
overall consequence of Hf-doping is that the CBM shifts up by
0.23 eV, while the VB maximum (VBM) remains more or less at
the same energy; the band gap at G increases to 3.34 eV.
Clearly, this does not go in the desired direction.
The presence of an element of lower valence in an oxide is
usually compensated by the formation of O vacancies (VO, in
case of Hf one per dopant). This leads to a reduced HfxW1 xO3 x
sample. Indeed, energy considerations show that while the removal of an O atom in the perfect crystal costs more than
5 eV, this value drops to zero and becomes even slightly negative for the Hf-doped material (VO formation energies computed with respect to 1=2 O2). Thermodynamically, doping WO3
with Hf will result in high concentrations of O vacancies. A VO
center in WO3 can be created along different directions, resulting in different electronic structures due to the anisotropic
nature of the material. Here we considered three cases where
VO is created along x, y, or z -W-O-W- chains, respectively. We
placed the VO and Hf defects at the largest possible distance in
the cell in order to avoid spurious interactions. The overall
atomic relaxation is larger than in absence of VO, with some O
atoms moving from the original position in the chain to an intermediate position between two different chains (see the Supporting Information). This large distortion has a non-negligible
effect on the WO6 octahedra and on the W 5d levels. Unlike
HfxW1 xO3, in HfxW1 xO3 x both VBM and CBM shift upwards
significantly. For VO along the x-chain (VO formation energy
DEf = 0.06 eV), the shifts are 0.63 and 0.24 eV, respectively
(see Table 1), leading to a reduced band gap of 2.71 eV. For VO
along the y-chain, the shifts are 0.53 and 0.76 eV, respectively,
Table 1 (band gap 3.33 eV, DEf = 0.35 eV). Finally, for VO along
the z-chain the band gap is 2.96 eV with VBM and CBM
upward shifts of 0.60 and 0.47 eV, respectively (DEf = 0.12 eV).
So, no matter along which direction the vacancy is created,
a substantial shift in both CBM and VBM is observed.
The upward shift of the bottom of the CB has been found
already in HfxW1 xO3. In HfxW1 xO3 x, however, the structural relaxation is larger, explaining the larger shift. Nevertheless, the
creation of an O vacancy has a direct and important effect also

 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemcatchem.org

477

Table 1. The total energy difference for HfxW1 xO3 x with VO along different -W-O-W- chains. Shifts of VBM and CBM for HfxW1 xO3 and
HfxW1 xO3 x.
Structure

Energy
[eV]

VBM shift
[eV]

CBM shift
[eV]

HfxW1 xO3
VO along x-chain
VO along y-chain
VO along z-chain

0
+0.41
+0.18

0
+0.63
+0.53
+0.60

+0.23
+0.24
+0.76
+0.47

on the position of the top of the VB, which shifts to higher energies, at variance with the case where no vacancies are present. The shift arises mainly from the O atoms located in the
same plane of Hf. In particular, VO introduces two extra electrons which fill the two O 2p hole states. These belong to O
atoms adjacent to the large Hf ion, resulting in an increased
Coulomb repulsion with consequent destabilization of the 2p
levels which move to higher energy thus shifting the top of
the VB.
To summarize, we have shown that a single dopant (Hf) replacing W in monoclinic WO3 generates the presence of a compensating defect (VO) and that this induces a shift of both the
top of the VB and the bottom of the CB towards higher energies, as schematically shown in Figure 4. This is a desirable

tential (ECP) combined with a modified HayWadt double-zeta


basis set[19, 20] and Haywsc-411d31,[21] respectively. In some
cases spin-polarized calculations were performed. The lattice
parameters of g-monoclinic WO3 (space group P21/n) were
taken from our previous work.[22] During the optimization, the
lattice parameters were fixed and only the atoms were allowed
to relax. The irreducible Brillouin zone (BZ) was sampled using
at least 36 k points for the 32-atoms cell corresponding to
a dopant concentration of 12.5 %. The results have been
checked also with a larger 64-atoms supercell. The equilibrium
structure was determined by using a quasi-Newton algorithm
with a BroydenFletcherGoldfarbShanno (BFGS) Hessian updating scheme.[23] The band edge shifts were calculated taking
as a reference the mean value of the 1s core levels of the O
atoms not in direct contact with the dopant.

Acknowledgements
This work is supported by CARIPLO Foundation through an Advanced Materials Grant 2009. Regione Lombardia and CILEA Consortium, through a LISA Initiative (Laboratory for Interdisciplinary
Advanced Simulation), are also gratefully acknowledged
Keywords: density functional calculations doping electronic
structure transition metal oxides water splitting

Figure 4. The schematic electron energy levels of the metal (W and Hf) and
of O atoms in a) WO3, b) HfxW1 xO3, and c) HfxW1 xO3 x. Dashed lines represent top of the VB and bottom of the CB for the pure oxide. Red arrows indicate levels shifts from a) to b) and from b) to c).

effect, which improves the reducing potential of WO3 for hydrogen evolution in photoactivated water splitting. The shift is
not accompanied by an increase of the band gap, thus maintaining the vis-light absorption properties of WO3, and is not
attributed to localized states in the gap, which may act as recombination centers. This is because the bands shifts result
from a different contribution to the VB and the CB by atoms
belonging to different structural planes. Involving a large
number of atoms, these are long range effects which affect the
whole band states.

Theoretical Section
We performed DFT calculations using the B3LYP[14, 15] hybrid
functional and the linear combination of atomic orbitals
(LCAO) approach as implemented in CRYSTAL09.[16, 17] The allelectron Gaussian-type basis set 8-411(d1) was adopted for
O,[18] while for W and Hf we used an effective core pseudopo-

478

www.chemcatchem.org

[1] A. Kudo, Y. Miseki, Chem. Soc. Rev. 2009, 38, 253.


[2] Special Issue in Chem. Phys.: Doping and functionalization of photoactive semiconducting oxides 2007, 339, pp. 1 192. Guest Editors: C.
Di Valentin, U. Diebold, A. Selloni.
[3] C. Di Valentin, G. Pacchioni, Catal. Today. 2011, DOI: 10.1016/
j.cattod.2011.11.030.
[4] K. Maeda, K. Domen, J. Phys. Chem. C 2007, 111, 7851.
[5] X. Chang, S. Sun, Y. Zhou, L. Dong, Y. Yin, Nanotechnology 2011, 22,
265603.
[6] M. N. Huda, Y. Yan, C.-Y. Moon, S.-H. Wei, M. M. Al-Jassim, Phys. Rev. B
2008, 77, 195 102.
[7] D. W. Hwang, J. Kim, T. J. Park, J. S. Lee, Catal. Lett. 2002, 80, 53.
[8] L. Zhou, J. Zhu, M. Yu, X. Huang, Z. Li, Y. Wang, C. Yu, J. Phys. Chem. C
2010, 114, 20947.
[9] A. Hameed, M. A. Gondal, Z. H. Yamani, Catal. Commun. 2004, 5, 715.
[10] B. Yang, V. Luca, Chem. Commun. 2008, 4454 4456.
[11] K. M. Karuppasamy, A. Subrahmanyam, J. Phys. D 2008, 41, 035 302.
[12] Y.-C. Nah, I. Paramasivam, R. Hahn, N. K. Shrestha, P. Schmuki, Nanotechnology 2010, 21, 105704.
[13] F. Wang, C. Di Valentin, G. Pacchioni, Phys. Rev. B 2011, 84, 073103.
[14] A. D. Becke, J. Chem. Phys. 1993, 98, 5648.
[15] C. Lee, W. Yang, R. G. Parr, Phys. Rev. B 1988, 37, 785.
[16] R. Dovesi, R. Orlando, B. Civalleri, C. Roetti, V. R. Saunders, C. M. Zicovich-Wilson, Z. Kristallogr. 2005, 220, 571.
[17] R. Dovesi et al., CRYSTAL09 Users Manual; University of Torino, Torino,
Italy, 2009.
[18] E. Ruiz, M. Llunell, P. Alemany, J. Solid State Chem. 2003, 176, 400.
[19] P. J. Hay, W. R. Wadt, J. Chem. Phys. 1985, 82, 270, 284, 299.
[20] P. H. Durand, J. C. Barthelat, Theor. Chim. Acta 1975, 38, 283.
[21] D. Munoz-Ramo, J. L. Gavartin, A. L. Shluger, Phys. Rev. B 2007, 75,
205 336.
[22] F. Wang, C. Di Valentin, G. Pacchioni, J. Phys. Chem. C 2011, 115, 8345.
[23] B. Civalleri, Ph. DArco, R. Orlando, V. R. Saunders, R. Dovesi, Chem. Phys.
Lett. 2001, 348, 131.
Received: November 29, 2011
Published online on March 1, 2012

 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ChemCatChem 2012, 4, 476 478

Potrebbero piacerti anche